AI may fail just like the music industry did

Centerpiece

Inconceivable Member
Messages
3,410
Here’s ya boy, Rick Beato laying out his thinking about how AI will fail, SOONER than later. And how some of these huge AI data centers will be sitting empty and unused. Not sure if I’m understanding it all as yet, but I’m a Luddite anyway. Check out this 9 minute vid which just dropped a couple of hours ago:
 
Here’s ya boy, Rick Beato laying out his thinking about how AI will fail, SOONER than later. And how some of these huge AI data centers will be sitting empty and unused. Not sure if I’m understanding it all as yet, but I’m a Luddite anyway. Check out this 9 minute vid which just dropped a couple of hours ago:

In watching that, I dont think he is saying that AI will fail so much as he is saying that the centralized version of AI servers we have come to know...will fail as a business model.

To me, he is essentially making the argument of Prodigy and AOL. Those were once the cutting edge of what we knew as online connection and offered centralized access in a one stop shop. However, they proved to be cumbersome, expensive, and inflexible once everyone got into the game of content creation and users only needed raw access.

In this comparison, I agree with him. The push will be for higher end GPU home computers that will process and maintain the preferred model of the user while updating their knowledge base from outside sources.

Think of it as deciding whether to use Netscape or Internet Explorer on your old machine once you had connection and access to rhe outside internet.
 
I hope all of it fails in spectacular fashion and takes every investment ever made into AI to zero. I honestly do.
Agreed and it can't happen fast enough.

I hate every single LLM. I use them for some things like writing emails and simple calculations but they all suck. They're nothing more than an algorithmic predictive chat bots yet people think they're actually having a conversation with them. I actually refuse to call them AI bc they're not.
 
Agreed and it can't happen fast enough.

I hate every single LLM. I use them for some things like writing emails and simple calculations but they all suck. They're nothing more than an algorithmic predictive chat bots yet people think they're actually having a conversation with them. I actually refuse to call them AI bc they're not.
I AGREE! Those folks who think their feces don't stink just because they have a LLM in addition to a plain old JD are absolutely insufferable.
 
Tldr, he's not arguing that AI will die. He's arguing that companies investing billions in Big Data centers to serve their models will never see the demand anticipated because it will be replaced by open source/free models that run on local machines.

He's essentially making the argument of AWS versus on premises data centers. Back in the 2000s, every 200 person company was building out their own data center with servers that were underutilized plus really expensive networking equipment. It was a massive expense for the hardware plus you had to pay the people to run them.

Then Amazon (and VMware) came along and said you could rent a small portion of thousands of their servers in a data center and it would be much cheaper because you're sharing the cost of the hardware, utilizing the equipment much more efficiently and you're sharing the cost of people to maintain the equipment.

But thats not the whole story. While it makes good business sense to run business apps like your email and your CRM at a central location, companies and individuals still pay beaucoup dollars to give each of their employees a computer with plenty of processing power.

He is saying that the open source models that are available today will be good enough for most people and businesses and they can just run them on local machines like a laptop, and the more powerful models that require more computing processing are just not needed and won't be adopted at the rate these companies are thinking.

He is definitely not saying AI will die because it's not useful. He's saying that the massive investment that some of these companies are making in a centralized AI model will not pay off and that the models that can run on a local machine will be what everyone uses.

Not sure if he's right. He was using qwen (made by ali baba) which is a good model but you have to trust the chinese government and you need to understand that at some point, they are probably going to try to sell you something. Deepseek is another Chinese company which is even more shadowy but they also produce an open source model. Mistral (my personal favorite open source) is a French company that produces a good open source model but it is not as good as Google or anthropic. Facebook/meta was going to be the leader in open source models but they look to have dramatically scaled back their investment and haven't released a new model in almost a year. I don't think they can see the path to profitability and after burning a whole lot of money on their VR stuff, I think they're a little gun shy to invest billions to give away something for free.
 
Last edited:
Will listen to Rick's take on music. Not sure this opinion is hitting the mark though.

The VAST majority of data center ai capacity is not being used to generate pantry recipes for the single mom in the burbs. They're being used by businesses and sovereign nations for a wide array of back-end and customer-facing applications. Yes, the recording industry became decentralized when the technology was miniaturized. This also happened to the television industry when camera studios were shrunk to the size of a mobile phone. But I think the data center industry will turn out more like the power company. We don't have a miniature natural-gas run turbine power plant for each house on the block. Its best kept on the edge of town and a utility fee to access their much larger grid is paid by us and (and much more importantly) large corporations.
 
Last edited:
Wait, someone hates the Beatles? I honestly am completely speechless at the thought.
For a regulatory policy course I taught, I used a lot of student interactions to illustrate policy issues and solutions. One of them involved the Beatles -- specifically, groups had to guess whether their group had more Beatles fans than the others. I picked that because it is probably not something most of them know about each other.

I was surprised that about 75% of the students did not like the Beatles.

Anecdotally, my son can't stand the Beatles; neither can his girlfriend (I had no formative role for her). My brothers do not like the Beatles. I mean, did you seriously think everyone would like the Beatles? Seriously?

The Beatles are in part a litmus test of what you think music should be doing. If you think that rock and roll should basically be pop (not today's pop but what was called pop, say, pre-1990), then the Beatles are good. If you have a different conception, then maybe the Beatles blow.
 
For a regulatory policy course I taught, I used a lot of student interactions to illustrate policy issues and solutions. One of them involved the Beatles -- specifically, groups had to guess whether their group had more Beatles fans than the others. I picked that because it is probably not something most of them know about each other.

I was surprised that about 75% of the students did not like the Beatles.

Anecdotally, my son can't stand the Beatles; neither can his girlfriend (I had no formative role for her). My brothers do not like the Beatles. I mean, did you seriously think everyone would like the Beatles? Seriously?

The Beatles are in part a litmus test of what you think music should be doing. If you think that rock and roll should basically be pop (not today's pop but what was called pop, say, pre-1990), then the Beatles are good. If you have a different conception, then maybe the Beatles blow.
I wonder how much of the Beatles catalog you and they have heard. Fwiw, my tastes have moved on but I still appreciate much of their music. I do question how anyone can think of them as merely pop, nttawwt.
 
Agreed and it can't happen fast enough.

I hate every single LLM. I use them for some things like writing emails and simple calculations but they all suck. They're nothing more than an algorithmic predictive chat bots yet people think they're actually having a conversation with them. I actually refuse to call them AI bc they're not.
mediocrity machines. their core programming is to regurgitate the most average example of their database, but with a people pleasing package.
 
Back
Top